A little background on my question.
Basically I want people to be able to say something like
prog page*.html morebookmarks.html
For Unix users, I want to be lazy and have the shell glob that for me. For
Windows users, I want to be lazy and have the perl function glob that for
me.
Q1. Could the redundant perl glob be harmful to files that have already been
shell globbed in Unix?
Q2. If bash globs page*.html and there are no matching files, is that
automatically an error even if there are other files are on the command line
to be processed? Or does bash consider you to be specifying zero or more
files that match page*.html and silently glob it to ''; I have been silently
hoping it is the second, but I am not sure.
Here is my test script:
print "Before perl globbing:\n ";
for $arg (@ARGV) {print "- $arg\n";};
my @Robs_way = map glob, @ARGV;
print "Rob\'s way:\n";
for $arg (@Robs_way) {print "- $arg\n";};
my @Johns_way = map glob, map { s/(\s)/\\$1/g; $_ } @ARGV;
print "John\'s way:\n";
for $arg (@Johns_way) {print "- $arg\n";};
Here is my command line and output:
C:\junk>testargs.pl "C:\Documents and Settings\deason\My Documents\xyz\junk
test
.html" *.html
Before perl globbing:
- C:\Documents and Settings\deason\My Documents\xyz\junk test.html
- *.html
Rob's way:
- C:./Documents
- and
- SettingsdeasonMy
- Documentsxyzjunk
- test.html
- junk3.html
- px600.html
John's way:
- C:./Documents and SettingsdeasonMy Documentsxyzjunk test.html
- junk3.html
- px600.html
Looks like I have to try Jenda's way!
"Jenda Krynicky" <[EMAIL PROTECTED]> wrote in message
3E1380AA.23905.3D8B65B0@localhost">news:3E1380AA.23905.3D8B65B0@localhost...
> From: "Rob Dixon" <[EMAIL PROTECTED]>
> > "Jenda Krynicky" <[EMAIL PROTECTED]> wrote in message
> > 3E1349E5.27703.3CB576EE@localhost">news:3E1349E5.27703.3CB576EE@localhost...
> > > From: "Rob Dixon" <[EMAIL PROTECTED]>
> > > > But glob 'absent_file.txt' returns ('absent_file.txt') so I think
> > > > this does what's required. As perldoc says, glob EXPR returns the
> > > > value of EXPR with filename expansions such as the standard Unix
> > > > shell /bin/csh would do.
> > >
> > > I see ... it only returns nothing if the parameter contained a * or
> > > ?
> > >
> > > I remember having some discussions regarding this and we agreed that
> > > even this should survive the cmdline globbing.
> >
> > Sounds like a little overactive ego on the part of the shell in that
> > case. You're saying that the script in
> >
> > script.pl *.ext file.ext
> >
> > would see an @ARGV of
> >
> > ('*.ext', 'file.ext')
> >
> > if there were no files with this extension, but
> >
> > ('file.ext', 'file1.ext', 'file2.ext', 'file.ext')
> >
> > if all of these (three) files existed?
> >
> > Then how do you pass '*.ext' as ('*.ext') if there are such files?
>
> If we are talking about unix shells
>
> script.pl '*.ext' file.ext
>
> And actually that's what will work with G.pm as well.
>
> And yes you are right, that's exactly what is the shell supposed to
> be doing.
>
> > I
> > suppose you can pick up the entire command line (I think?) and process
> > that, but I'm very glad most languages don't do clever 'useful' things
> > with your parameter list before called code gets to see them.
>
> The whole problem is that the process gets the commandline if you use
> Windows and a list of parameters if you use Unix.
>
> Normaly the C runtime under Windows parses the commandline and fills
> the argv array. Which is exactly what happens if you run a Perl
> script under windows. The system passes the cmdline and the runtime
> breaks it apart. And breaks it. And your program is supposed to do
> the globbing, even though the runtime already stripped some well
> needed info.
>
> The G.pm basicaly overcomes the broken @ARGV and parses the cmdline
> and globs the parameters at the same time. Close to the Unix way.
>
> If you do not want that you either do not use G.pm and glob just the
> stuff you want or
> use G qw(NOGLOB StripPerlFromCommandLine GetCommandLine);
>
> $rawcmdline = GetCommandLine();
> $cmdlineparams = StripPerlFromCommandLine($rawcmdline);
>
> and parse the raw commandline.
> (Actually it's not totally raw. The IO redirections and pipes are
> already processed by the shell and NOT passed to the process.)
>
> > > So that you could print
> > >
> > > Cannot open file *.txt
> > > or
> > > No files match the *.txt mask
> > >
> > >
> > > If you happen to have several masks with different meanings it might
> > > be good to be able to say which one did not return anything. And it
> > > may not of course :-) I was told this is the behaviour of the Unix
> > > shells ...
> > >
> >
> > Fine, but rather than have the shell performing obscure gymnastics
> > beneath me I'd rather code it explicitly:
> >
> > for (@ARGV) { print "No files match the $_ mask\n" unless glob };
> > my @input_files = map glob, @ARGV;
>
> Problem is that under Unix the shell IS performing the gymnastics.
>
> You can quote some params to ensure they are not globbed, but the
> default is globbing.
>
> > > > > From: "Rob Dixon" <[EMAIL PROTECTED]>
> > > > > > my @input_files = map glob, @ARGV;
> > >
> > > There's one more bad thing about this code. Suppose the script was
> > > called like this
> > >
> > > script.pl "hello world.doc"
> > >
> > > then the @ARGV = ( 'hello world')
> > > but the line above will set @input_files to
> > > ( 'hello', 'world.doc' )
> >
> > Not at all. @ARGV would equal ( "hello world.doc" ) and the statement
> > would set @input_files = ( "hello world.doc" ).
>
> What's your version of Perl?
>
> In file zkGlob.pl I have this:
> #!perl
> print "\@ARGV = ( '" . join( "', '", @ARGV) . "')\n\n";
> @input_files = map {glob $_} @ARGV;
> print "\@input_files = ( '" . join( "', '", @input_files) .
> "')\n\n";
>
> call it as:
> c:\temp\zkGlob.pl "hello world.doc" *.txt
>
> and get:
> @ARGV = ( 'hello world.doc', '*.txt')
>
> @input_files = ( 'hello', 'world.doc', 'te-fetch.txt')
>
> Jenda
> ===== [EMAIL PROTECTED] === http://Jenda.Krynicky.cz =====
> When it comes to wine, women and song, wizards are allowed
> to get drunk and croon as much as they like.
> -- Terry Pratchett in Sourcery
>
--
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
